2 * Copyright (c) 2000-2006 Silicon Graphics, Inc.
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
20 #include "xfs_shared.h"
21 #include "xfs_format.h"
22 #include "xfs_log_format.h"
23 #include "xfs_trans_resv.h"
25 #include "xfs_mount.h"
26 #include "xfs_da_format.h"
27 #include "xfs_inode.h"
28 #include "xfs_btree.h"
30 #include "xfs_alloc.h"
31 #include "xfs_error.h"
32 #include "xfs_fsops.h"
33 #include "xfs_trans.h"
34 #include "xfs_buf_item.h"
36 #include "xfs_log_priv.h"
37 #include "xfs_da_btree.h"
39 #include "xfs_extfree_item.h"
40 #include "xfs_mru_cache.h"
41 #include "xfs_inode_item.h"
42 #include "xfs_icache.h"
43 #include "xfs_trace.h"
44 #include "xfs_icreate_item.h"
45 #include "xfs_filestream.h"
46 #include "xfs_quota.h"
47 #include "xfs_sysfs.h"
49 #include <linux/namei.h>
50 #include <linux/init.h>
51 #include <linux/slab.h>
52 #include <linux/mount.h>
53 #include <linux/mempool.h>
54 #include <linux/writeback.h>
55 #include <linux/kthread.h>
56 #include <linux/freezer.h>
57 #include <linux/parser.h>
59 static const struct super_operations xfs_super_operations;
60 static kmem_zone_t *xfs_ioend_zone;
61 mempool_t *xfs_ioend_pool;
63 static struct kset *xfs_kset; /* top-level xfs sysfs dir */
65 static struct xfs_kobj xfs_dbg_kobj; /* global debug sysfs attrs */
68 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
69 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
70 #define MNTOPT_LOGDEV "logdev" /* log device */
71 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
72 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
73 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
74 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
75 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
76 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
77 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
78 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
79 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
80 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
81 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
82 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
83 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
84 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
85 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
86 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
87 * unwritten extent conversion */
88 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
89 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
90 #define MNTOPT_32BITINODE "inode32" /* inode allocation limited to
91 * XFS_MAXINUMBER_32 */
92 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
93 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
94 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
95 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
97 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
98 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
99 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
100 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
101 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
102 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
103 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
104 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
105 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
106 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
107 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
108 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
109 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
110 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
111 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
112 #define MNTOPT_DELAYLOG "delaylog" /* Delayed logging enabled */
113 #define MNTOPT_NODELAYLOG "nodelaylog" /* Delayed logging disabled */
114 #define MNTOPT_DISCARD "discard" /* Discard unused blocks */
115 #define MNTOPT_NODISCARD "nodiscard" /* Do not discard unused blocks */
118 * Table driven mount option parser.
120 * Currently only used for remount, but it will be used for mount
121 * in the future, too.
131 static const match_table_t tokens = {
132 {Opt_barrier, "barrier"},
133 {Opt_nobarrier, "nobarrier"},
134 {Opt_inode64, "inode64"},
135 {Opt_inode32, "inode32"},
141 suffix_kstrtoint(char *s, unsigned int base, int *res)
143 int last, shift_left_factor = 0, _res;
146 last = strlen(value) - 1;
147 if (value[last] == 'K' || value[last] == 'k') {
148 shift_left_factor = 10;
151 if (value[last] == 'M' || value[last] == 'm') {
152 shift_left_factor = 20;
155 if (value[last] == 'G' || value[last] == 'g') {
156 shift_left_factor = 30;
160 if (kstrtoint(s, base, &_res))
162 *res = _res << shift_left_factor;
167 * This function fills in xfs_mount_t fields based on mount args.
168 * Note: the superblock has _not_ yet been read in.
170 * Note that this function leaks the various device name allocations on
171 * failure. The caller takes care of them.
175 struct xfs_mount *mp,
178 struct super_block *sb = mp->m_super;
179 char *this_char, *value;
183 __uint8_t iosizelog = 0;
186 * set up the mount name first so all the errors will refer to the
189 mp->m_fsname = kstrndup(sb->s_id, MAXNAMELEN, GFP_KERNEL);
192 mp->m_fsname_len = strlen(mp->m_fsname) + 1;
195 * Copy binary VFS mount flags we are interested in.
197 if (sb->s_flags & MS_RDONLY)
198 mp->m_flags |= XFS_MOUNT_RDONLY;
199 if (sb->s_flags & MS_DIRSYNC)
200 mp->m_flags |= XFS_MOUNT_DIRSYNC;
201 if (sb->s_flags & MS_SYNCHRONOUS)
202 mp->m_flags |= XFS_MOUNT_WSYNC;
205 * Set some default flags that could be cleared by the mount option
208 mp->m_flags |= XFS_MOUNT_BARRIER;
209 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
212 * These can be overridden by the mount option parsing.
220 while ((this_char = strsep(&options, ",")) != NULL) {
223 if ((value = strchr(this_char, '=')) != NULL)
226 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
227 if (!value || !*value) {
228 xfs_warn(mp, "%s option requires an argument",
232 if (kstrtoint(value, 10, &mp->m_logbufs))
234 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
235 if (!value || !*value) {
236 xfs_warn(mp, "%s option requires an argument",
240 if (suffix_kstrtoint(value, 10, &mp->m_logbsize))
242 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
243 if (!value || !*value) {
244 xfs_warn(mp, "%s option requires an argument",
248 mp->m_logname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
251 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
252 xfs_warn(mp, "%s option not allowed on this system",
255 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
256 if (!value || !*value) {
257 xfs_warn(mp, "%s option requires an argument",
261 mp->m_rtname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
264 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
265 if (!value || !*value) {
266 xfs_warn(mp, "%s option requires an argument",
270 if (kstrtoint(value, 10, &iosize))
272 iosizelog = ffs(iosize) - 1;
273 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
274 if (!value || !*value) {
275 xfs_warn(mp, "%s option requires an argument",
279 if (suffix_kstrtoint(value, 10, &iosize))
281 iosizelog = ffs(iosize) - 1;
282 } else if (!strcmp(this_char, MNTOPT_GRPID) ||
283 !strcmp(this_char, MNTOPT_BSDGROUPS)) {
284 mp->m_flags |= XFS_MOUNT_GRPID;
285 } else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
286 !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
287 mp->m_flags &= ~XFS_MOUNT_GRPID;
288 } else if (!strcmp(this_char, MNTOPT_WSYNC)) {
289 mp->m_flags |= XFS_MOUNT_WSYNC;
290 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
291 mp->m_flags |= XFS_MOUNT_NORECOVERY;
292 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
293 mp->m_flags |= XFS_MOUNT_NOALIGN;
294 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
295 mp->m_flags |= XFS_MOUNT_SWALLOC;
296 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
297 if (!value || !*value) {
298 xfs_warn(mp, "%s option requires an argument",
302 if (kstrtoint(value, 10, &dsunit))
304 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
305 if (!value || !*value) {
306 xfs_warn(mp, "%s option requires an argument",
310 if (kstrtoint(value, 10, &dswidth))
312 } else if (!strcmp(this_char, MNTOPT_32BITINODE)) {
313 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
314 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
315 mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS;
316 } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
317 mp->m_flags |= XFS_MOUNT_NOUUID;
318 } else if (!strcmp(this_char, MNTOPT_BARRIER)) {
319 mp->m_flags |= XFS_MOUNT_BARRIER;
320 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
321 mp->m_flags &= ~XFS_MOUNT_BARRIER;
322 } else if (!strcmp(this_char, MNTOPT_IKEEP)) {
323 mp->m_flags |= XFS_MOUNT_IKEEP;
324 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
325 mp->m_flags &= ~XFS_MOUNT_IKEEP;
326 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
327 mp->m_flags &= ~XFS_MOUNT_COMPAT_IOSIZE;
328 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
329 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
330 } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
331 mp->m_flags |= XFS_MOUNT_ATTR2;
332 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
333 mp->m_flags &= ~XFS_MOUNT_ATTR2;
334 mp->m_flags |= XFS_MOUNT_NOATTR2;
335 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
336 mp->m_flags |= XFS_MOUNT_FILESTREAMS;
337 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
338 mp->m_qflags &= ~XFS_ALL_QUOTA_ACCT;
339 mp->m_qflags &= ~XFS_ALL_QUOTA_ENFD;
340 mp->m_qflags &= ~XFS_ALL_QUOTA_ACTIVE;
341 } else if (!strcmp(this_char, MNTOPT_QUOTA) ||
342 !strcmp(this_char, MNTOPT_UQUOTA) ||
343 !strcmp(this_char, MNTOPT_USRQUOTA)) {
344 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
346 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
347 !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
348 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
349 mp->m_qflags &= ~XFS_UQUOTA_ENFD;
350 } else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
351 !strcmp(this_char, MNTOPT_PRJQUOTA)) {
352 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
354 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
355 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
356 mp->m_qflags &= ~XFS_PQUOTA_ENFD;
357 } else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
358 !strcmp(this_char, MNTOPT_GRPQUOTA)) {
359 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
361 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
362 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
363 mp->m_qflags &= ~XFS_GQUOTA_ENFD;
364 } else if (!strcmp(this_char, MNTOPT_DELAYLOG)) {
366 "delaylog is the default now, option is deprecated.");
367 } else if (!strcmp(this_char, MNTOPT_NODELAYLOG)) {
369 "nodelaylog support has been removed, option is deprecated.");
370 } else if (!strcmp(this_char, MNTOPT_DISCARD)) {
371 mp->m_flags |= XFS_MOUNT_DISCARD;
372 } else if (!strcmp(this_char, MNTOPT_NODISCARD)) {
373 mp->m_flags &= ~XFS_MOUNT_DISCARD;
374 } else if (!strcmp(this_char, "ihashsize")) {
376 "ihashsize no longer used, option is deprecated.");
377 } else if (!strcmp(this_char, "osyncisdsync")) {
379 "osyncisdsync has no effect, option is deprecated.");
380 } else if (!strcmp(this_char, "osyncisosync")) {
382 "osyncisosync has no effect, option is deprecated.");
383 } else if (!strcmp(this_char, "irixsgid")) {
385 "irixsgid is now a sysctl(2) variable, option is deprecated.");
387 xfs_warn(mp, "unknown mount option [%s].", this_char);
393 * no recovery flag requires a read-only mount
395 if ((mp->m_flags & XFS_MOUNT_NORECOVERY) &&
396 !(mp->m_flags & XFS_MOUNT_RDONLY)) {
397 xfs_warn(mp, "no-recovery mounts must be read-only.");
401 if ((mp->m_flags & XFS_MOUNT_NOALIGN) && (dsunit || dswidth)) {
403 "sunit and swidth options incompatible with the noalign option");
407 #ifndef CONFIG_XFS_QUOTA
408 if (XFS_IS_QUOTA_RUNNING(mp)) {
409 xfs_warn(mp, "quota support not available in this kernel.");
414 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
415 xfs_warn(mp, "sunit and swidth must be specified together");
419 if (dsunit && (dswidth % dsunit != 0)) {
421 "stripe width (%d) must be a multiple of the stripe unit (%d)",
427 if (dsunit && !(mp->m_flags & XFS_MOUNT_NOALIGN)) {
429 * At this point the superblock has not been read
430 * in, therefore we do not know the block size.
431 * Before the mount call ends we will convert
434 mp->m_dalign = dsunit;
435 mp->m_swidth = dswidth;
438 if (mp->m_logbufs != -1 &&
439 mp->m_logbufs != 0 &&
440 (mp->m_logbufs < XLOG_MIN_ICLOGS ||
441 mp->m_logbufs > XLOG_MAX_ICLOGS)) {
442 xfs_warn(mp, "invalid logbufs value: %d [not %d-%d]",
443 mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
446 if (mp->m_logbsize != -1 &&
447 mp->m_logbsize != 0 &&
448 (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE ||
449 mp->m_logbsize > XLOG_MAX_RECORD_BSIZE ||
450 !is_power_of_2(mp->m_logbsize))) {
452 "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
458 if (iosizelog > XFS_MAX_IO_LOG ||
459 iosizelog < XFS_MIN_IO_LOG) {
460 xfs_warn(mp, "invalid log iosize: %d [not %d-%d]",
461 iosizelog, XFS_MIN_IO_LOG,
466 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
467 mp->m_readio_log = iosizelog;
468 mp->m_writeio_log = iosizelog;
474 struct proc_xfs_info {
481 struct xfs_mount *mp,
484 static struct proc_xfs_info xfs_info_set[] = {
485 /* the few simple ones we can get from the mount struct */
486 { XFS_MOUNT_IKEEP, "," MNTOPT_IKEEP },
487 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
488 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
489 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
490 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
491 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
492 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 },
493 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
494 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
495 { XFS_MOUNT_DISCARD, "," MNTOPT_DISCARD },
496 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_32BITINODE },
499 static struct proc_xfs_info xfs_info_unset[] = {
500 /* the few simple ones we can get from the mount struct */
501 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO },
502 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER },
503 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE },
506 struct proc_xfs_info *xfs_infop;
508 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
509 if (mp->m_flags & xfs_infop->flag)
510 seq_puts(m, xfs_infop->str);
512 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
513 if (!(mp->m_flags & xfs_infop->flag))
514 seq_puts(m, xfs_infop->str);
517 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
518 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
519 (int)(1 << mp->m_writeio_log) >> 10);
521 if (mp->m_logbufs > 0)
522 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
523 if (mp->m_logbsize > 0)
524 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
527 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
529 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
531 if (mp->m_dalign > 0)
532 seq_printf(m, "," MNTOPT_SUNIT "=%d",
533 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
534 if (mp->m_swidth > 0)
535 seq_printf(m, "," MNTOPT_SWIDTH "=%d",
536 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
538 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
539 seq_puts(m, "," MNTOPT_USRQUOTA);
540 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
541 seq_puts(m, "," MNTOPT_UQUOTANOENF);
543 if (mp->m_qflags & XFS_PQUOTA_ACCT) {
544 if (mp->m_qflags & XFS_PQUOTA_ENFD)
545 seq_puts(m, "," MNTOPT_PRJQUOTA);
547 seq_puts(m, "," MNTOPT_PQUOTANOENF);
549 if (mp->m_qflags & XFS_GQUOTA_ACCT) {
550 if (mp->m_qflags & XFS_GQUOTA_ENFD)
551 seq_puts(m, "," MNTOPT_GRPQUOTA);
553 seq_puts(m, "," MNTOPT_GQUOTANOENF);
556 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
557 seq_puts(m, "," MNTOPT_NOQUOTA);
563 unsigned int blockshift)
565 unsigned int pagefactor = 1;
566 unsigned int bitshift = BITS_PER_LONG - 1;
568 /* Figure out maximum filesize, on Linux this can depend on
569 * the filesystem blocksize (on 32 bit platforms).
570 * __block_write_begin does this in an [unsigned] long...
571 * page->index << (PAGE_CACHE_SHIFT - bbits)
572 * So, for page sized blocks (4K on 32 bit platforms),
573 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
574 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
575 * but for smaller blocksizes it is less (bbits = log2 bsize).
576 * Note1: get_block_t takes a long (implicit cast from above)
577 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
578 * can optionally convert the [unsigned] long from above into
579 * an [unsigned] long long.
582 #if BITS_PER_LONG == 32
583 # if defined(CONFIG_LBDAF)
584 ASSERT(sizeof(sector_t) == 8);
585 pagefactor = PAGE_CACHE_SIZE;
586 bitshift = BITS_PER_LONG;
588 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
592 return (((__uint64_t)pagefactor) << bitshift) - 1;
596 * xfs_set_inode32() and xfs_set_inode64() are passed an agcount
597 * because in the growfs case, mp->m_sb.sb_agcount is not updated
598 * yet to the potentially higher ag count.
601 xfs_set_inode32(struct xfs_mount *mp, xfs_agnumber_t agcount)
603 xfs_agnumber_t index = 0;
604 xfs_agnumber_t maxagi = 0;
605 xfs_sb_t *sbp = &mp->m_sb;
606 xfs_agnumber_t max_metadata;
611 /* Calculate how much should be reserved for inodes to meet
612 * the max inode percentage.
614 if (mp->m_maxicount) {
617 icount = sbp->sb_dblocks * sbp->sb_imax_pct;
619 icount += sbp->sb_agblocks - 1;
620 do_div(icount, sbp->sb_agblocks);
621 max_metadata = icount;
623 max_metadata = agcount;
626 agino = XFS_OFFBNO_TO_AGINO(mp, sbp->sb_agblocks - 1, 0);
628 for (index = 0; index < agcount; index++) {
629 ino = XFS_AGINO_TO_INO(mp, index, agino);
631 if (ino > XFS_MAXINUMBER_32) {
632 pag = xfs_perag_get(mp, index);
633 pag->pagi_inodeok = 0;
634 pag->pagf_metadata = 0;
639 pag = xfs_perag_get(mp, index);
640 pag->pagi_inodeok = 1;
642 if (index < max_metadata)
643 pag->pagf_metadata = 1;
646 mp->m_flags |= (XFS_MOUNT_32BITINODES |
647 XFS_MOUNT_SMALL_INUMS);
653 xfs_set_inode64(struct xfs_mount *mp, xfs_agnumber_t agcount)
655 xfs_agnumber_t index = 0;
657 for (index = 0; index < agcount; index++) {
658 struct xfs_perag *pag;
660 pag = xfs_perag_get(mp, index);
661 pag->pagi_inodeok = 1;
662 pag->pagf_metadata = 0;
666 /* There is no need for lock protection on m_flags,
667 * the rw_semaphore of the VFS superblock is locked
668 * during mount/umount/remount operations, so this is
669 * enough to avoid concurency on the m_flags field
671 mp->m_flags &= ~(XFS_MOUNT_32BITINODES |
672 XFS_MOUNT_SMALL_INUMS);
680 struct block_device **bdevp)
684 *bdevp = blkdev_get_by_path(name, FMODE_READ|FMODE_WRITE|FMODE_EXCL,
686 if (IS_ERR(*bdevp)) {
687 error = PTR_ERR(*bdevp);
688 xfs_warn(mp, "Invalid device [%s], error=%d", name, error);
696 struct block_device *bdev)
699 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
703 xfs_blkdev_issue_flush(
704 xfs_buftarg_t *buftarg)
706 blkdev_issue_flush(buftarg->bt_bdev, GFP_NOFS, NULL);
711 struct xfs_mount *mp)
713 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
714 struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
715 xfs_free_buftarg(mp, mp->m_logdev_targp);
716 xfs_blkdev_put(logdev);
718 if (mp->m_rtdev_targp) {
719 struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
720 xfs_free_buftarg(mp, mp->m_rtdev_targp);
721 xfs_blkdev_put(rtdev);
723 xfs_free_buftarg(mp, mp->m_ddev_targp);
727 * The file system configurations are:
728 * (1) device (partition) with data and internal log
729 * (2) logical volume with data and log subvolumes.
730 * (3) logical volume with data, log, and realtime subvolumes.
732 * We only have to handle opening the log and realtime volumes here if
733 * they are present. The data subvolume has already been opened by
734 * get_sb_bdev() and is stored in sb->s_bdev.
738 struct xfs_mount *mp)
740 struct block_device *ddev = mp->m_super->s_bdev;
741 struct block_device *logdev = NULL, *rtdev = NULL;
745 * Open real time and log devices - order is important.
748 error = xfs_blkdev_get(mp, mp->m_logname, &logdev);
754 error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev);
756 goto out_close_logdev;
758 if (rtdev == ddev || rtdev == logdev) {
760 "Cannot mount filesystem with identical rtdev and ddev/logdev.");
762 goto out_close_rtdev;
767 * Setup xfs_mount buffer target pointers
770 mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev);
771 if (!mp->m_ddev_targp)
772 goto out_close_rtdev;
775 mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev);
776 if (!mp->m_rtdev_targp)
777 goto out_free_ddev_targ;
780 if (logdev && logdev != ddev) {
781 mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev);
782 if (!mp->m_logdev_targp)
783 goto out_free_rtdev_targ;
785 mp->m_logdev_targp = mp->m_ddev_targp;
791 if (mp->m_rtdev_targp)
792 xfs_free_buftarg(mp, mp->m_rtdev_targp);
794 xfs_free_buftarg(mp, mp->m_ddev_targp);
796 xfs_blkdev_put(rtdev);
798 if (logdev && logdev != ddev)
799 xfs_blkdev_put(logdev);
805 * Setup xfs_mount buffer target pointers based on superblock
809 struct xfs_mount *mp)
813 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_sectsize);
817 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
818 unsigned int log_sector_size = BBSIZE;
820 if (xfs_sb_version_hassector(&mp->m_sb))
821 log_sector_size = mp->m_sb.sb_logsectsize;
822 error = xfs_setsize_buftarg(mp->m_logdev_targp,
827 if (mp->m_rtdev_targp) {
828 error = xfs_setsize_buftarg(mp->m_rtdev_targp,
829 mp->m_sb.sb_sectsize);
838 xfs_init_mount_workqueues(
839 struct xfs_mount *mp)
841 mp->m_buf_workqueue = alloc_workqueue("xfs-buf/%s",
842 WQ_MEM_RECLAIM|WQ_FREEZABLE, 1, mp->m_fsname);
843 if (!mp->m_buf_workqueue)
846 mp->m_data_workqueue = alloc_workqueue("xfs-data/%s",
847 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
848 if (!mp->m_data_workqueue)
849 goto out_destroy_buf;
851 mp->m_unwritten_workqueue = alloc_workqueue("xfs-conv/%s",
852 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
853 if (!mp->m_unwritten_workqueue)
854 goto out_destroy_data_iodone_queue;
856 mp->m_cil_workqueue = alloc_workqueue("xfs-cil/%s",
857 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
858 if (!mp->m_cil_workqueue)
859 goto out_destroy_unwritten;
861 mp->m_reclaim_workqueue = alloc_workqueue("xfs-reclaim/%s",
862 WQ_FREEZABLE, 0, mp->m_fsname);
863 if (!mp->m_reclaim_workqueue)
864 goto out_destroy_cil;
866 mp->m_log_workqueue = alloc_workqueue("xfs-log/%s",
867 WQ_FREEZABLE|WQ_HIGHPRI, 0, mp->m_fsname);
868 if (!mp->m_log_workqueue)
869 goto out_destroy_reclaim;
871 mp->m_eofblocks_workqueue = alloc_workqueue("xfs-eofblocks/%s",
872 WQ_FREEZABLE, 0, mp->m_fsname);
873 if (!mp->m_eofblocks_workqueue)
874 goto out_destroy_log;
879 destroy_workqueue(mp->m_log_workqueue);
881 destroy_workqueue(mp->m_reclaim_workqueue);
883 destroy_workqueue(mp->m_cil_workqueue);
884 out_destroy_unwritten:
885 destroy_workqueue(mp->m_unwritten_workqueue);
886 out_destroy_data_iodone_queue:
887 destroy_workqueue(mp->m_data_workqueue);
889 destroy_workqueue(mp->m_buf_workqueue);
895 xfs_destroy_mount_workqueues(
896 struct xfs_mount *mp)
898 destroy_workqueue(mp->m_eofblocks_workqueue);
899 destroy_workqueue(mp->m_log_workqueue);
900 destroy_workqueue(mp->m_reclaim_workqueue);
901 destroy_workqueue(mp->m_cil_workqueue);
902 destroy_workqueue(mp->m_data_workqueue);
903 destroy_workqueue(mp->m_unwritten_workqueue);
904 destroy_workqueue(mp->m_buf_workqueue);
908 * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK
909 * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting
910 * for IO to complete so that we effectively throttle multiple callers to the
911 * rate at which IO is completing.
915 struct xfs_mount *mp)
917 struct super_block *sb = mp->m_super;
919 if (down_read_trylock(&sb->s_umount)) {
921 up_read(&sb->s_umount);
925 /* Catch misguided souls that try to use this interface on XFS */
926 STATIC struct inode *
928 struct super_block *sb)
935 * Now that the generic code is guaranteed not to be accessing
936 * the linux inode, we can reclaim the inode.
939 xfs_fs_destroy_inode(
942 struct xfs_inode *ip = XFS_I(inode);
944 trace_xfs_destroy_inode(ip);
946 XFS_STATS_INC(vn_reclaim);
948 ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0);
951 * We should never get here with one of the reclaim flags already set.
953 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIMABLE));
954 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIM));
957 * We always use background reclaim here because even if the
958 * inode is clean, it still may be under IO and hence we have
959 * to take the flush lock. The background reclaim path handles
960 * this more efficiently than we can here, so simply let background
961 * reclaim tear down all inodes.
963 xfs_inode_set_reclaim_tag(ip);
967 * Slab object creation initialisation for the XFS inode.
968 * This covers only the idempotent fields in the XFS inode;
969 * all other fields need to be initialised on allocation
970 * from the slab. This avoids the need to repeatedly initialise
971 * fields in the xfs inode that left in the initialise state
972 * when freeing the inode.
975 xfs_fs_inode_init_once(
978 struct xfs_inode *ip = inode;
980 memset(ip, 0, sizeof(struct xfs_inode));
983 inode_init_once(VFS_I(ip));
986 atomic_set(&ip->i_pincount, 0);
987 spin_lock_init(&ip->i_flags_lock);
989 mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
990 "xfsino", ip->i_ino);
997 xfs_inode_t *ip = XFS_I(inode);
999 ASSERT(!rwsem_is_locked(&ip->i_iolock.mr_lock));
1001 trace_xfs_evict_inode(ip);
1003 truncate_inode_pages_final(&inode->i_data);
1005 XFS_STATS_INC(vn_rele);
1006 XFS_STATS_INC(vn_remove);
1012 * We do an unlocked check for XFS_IDONTCACHE here because we are already
1013 * serialised against cache hits here via the inode->i_lock and igrab() in
1014 * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be
1015 * racing with us, and it avoids needing to grab a spinlock here for every inode
1016 * we drop the final reference on.
1020 struct inode *inode)
1022 struct xfs_inode *ip = XFS_I(inode);
1024 return generic_drop_inode(inode) || (ip->i_flags & XFS_IDONTCACHE);
1029 struct xfs_mount *mp)
1031 kfree(mp->m_fsname);
1032 kfree(mp->m_rtname);
1033 kfree(mp->m_logname);
1038 struct super_block *sb)
1040 struct xfs_mount *mp = XFS_M(sb);
1042 xfs_filestream_unmount(mp);
1046 xfs_icsb_destroy_counters(mp);
1047 xfs_destroy_mount_workqueues(mp);
1048 xfs_close_devices(mp);
1049 xfs_free_fsname(mp);
1055 struct super_block *sb,
1058 struct xfs_mount *mp = XFS_M(sb);
1061 * Doing anything during the async pass would be counterproductive.
1066 xfs_log_force(mp, XFS_LOG_SYNC);
1069 * The disk must be active because we're syncing.
1070 * We schedule log work now (now that the disk is
1071 * active) instead of later (when it might not be).
1073 flush_delayed_work(&mp->m_log->l_work);
1081 struct dentry *dentry,
1082 struct kstatfs *statp)
1084 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1085 xfs_sb_t *sbp = &mp->m_sb;
1086 struct xfs_inode *ip = XFS_I(dentry->d_inode);
1087 __uint64_t fakeinos, id;
1093 statp->f_type = XFS_SB_MAGIC;
1094 statp->f_namelen = MAXNAMELEN - 1;
1096 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1097 statp->f_fsid.val[0] = (u32)id;
1098 statp->f_fsid.val[1] = (u32)(id >> 32);
1100 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
1101 icount = percpu_counter_sum(&mp->m_icount);
1102 ifree = percpu_counter_sum(&mp->m_ifree);
1104 spin_lock(&mp->m_sb_lock);
1105 statp->f_bsize = sbp->sb_blocksize;
1106 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1107 statp->f_blocks = sbp->sb_dblocks - lsize;
1108 statp->f_bfree = statp->f_bavail =
1109 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1110 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1111 statp->f_files = MIN(icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1112 if (mp->m_maxicount)
1113 statp->f_files = min_t(typeof(statp->f_files),
1117 /* If sb_icount overshot maxicount, report actual allocation */
1118 statp->f_files = max_t(typeof(statp->f_files),
1122 /* make sure statp->f_ffree does not underflow */
1123 ffree = statp->f_files - (icount - ifree);
1124 statp->f_ffree = max_t(__int64_t, ffree, 0);
1126 spin_unlock(&mp->m_sb_lock);
1128 if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) &&
1129 ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))) ==
1130 (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))
1131 xfs_qm_statvfs(ip, statp);
1136 xfs_save_resvblks(struct xfs_mount *mp)
1138 __uint64_t resblks = 0;
1140 mp->m_resblks_save = mp->m_resblks;
1141 xfs_reserve_blocks(mp, &resblks, NULL);
1145 xfs_restore_resvblks(struct xfs_mount *mp)
1149 if (mp->m_resblks_save) {
1150 resblks = mp->m_resblks_save;
1151 mp->m_resblks_save = 0;
1153 resblks = xfs_default_resblks(mp);
1155 xfs_reserve_blocks(mp, &resblks, NULL);
1159 * Trigger writeback of all the dirty metadata in the file system.
1161 * This ensures that the metadata is written to their location on disk rather
1162 * than just existing in transactions in the log. This means after a quiesce
1163 * there is no log replay required to write the inodes to disk - this is the
1164 * primary difference between a sync and a quiesce.
1166 * Note: xfs_log_quiesce() stops background log work - the callers must ensure
1167 * it is started again when appropriate.
1171 struct xfs_mount *mp)
1175 /* wait for all modifications to complete */
1176 while (atomic_read(&mp->m_active_trans) > 0)
1179 /* force the log to unpin objects from the now complete transactions */
1180 xfs_log_force(mp, XFS_LOG_SYNC);
1182 /* reclaim inodes to do any IO before the freeze completes */
1183 xfs_reclaim_inodes(mp, 0);
1184 xfs_reclaim_inodes(mp, SYNC_WAIT);
1186 /* Push the superblock and write an unmount record */
1187 error = xfs_log_sbcount(mp);
1189 xfs_warn(mp, "xfs_attr_quiesce: failed to log sb changes. "
1190 "Frozen image may not be consistent.");
1192 * Just warn here till VFS can correctly support
1193 * read-only remount without racing.
1195 WARN_ON(atomic_read(&mp->m_active_trans) != 0);
1197 xfs_log_quiesce(mp);
1202 struct super_block *sb,
1206 struct xfs_mount *mp = XFS_M(sb);
1207 xfs_sb_t *sbp = &mp->m_sb;
1208 substring_t args[MAX_OPT_ARGS];
1212 sync_filesystem(sb);
1213 while ((p = strsep(&options, ",")) != NULL) {
1219 token = match_token(p, tokens, args);
1222 mp->m_flags |= XFS_MOUNT_BARRIER;
1225 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1228 mp->m_maxagi = xfs_set_inode64(mp, sbp->sb_agcount);
1231 mp->m_maxagi = xfs_set_inode32(mp, sbp->sb_agcount);
1235 * Logically we would return an error here to prevent
1236 * users from believing they might have changed
1237 * mount options using remount which can't be changed.
1239 * But unfortunately mount(8) adds all options from
1240 * mtab and fstab to the mount arguments in some cases
1241 * so we can't blindly reject options, but have to
1242 * check for each specified option if it actually
1243 * differs from the currently set option and only
1244 * reject it if that's the case.
1246 * Until that is implemented we return success for
1247 * every remount request, and silently ignore all
1248 * options that we can't actually change.
1252 "mount option \"%s\" not supported for remount", p);
1261 if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) {
1262 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1265 * If this is the first remount to writeable state we
1266 * might have some superblock changes to update.
1268 if (mp->m_update_sb) {
1269 error = xfs_sync_sb(mp, false);
1271 xfs_warn(mp, "failed to write sb changes");
1274 mp->m_update_sb = false;
1278 * Fill out the reserve pool if it is empty. Use the stashed
1279 * value if it is non-zero, otherwise go with the default.
1281 xfs_restore_resvblks(mp);
1282 xfs_log_work_queue(mp);
1286 if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
1288 * Before we sync the metadata, we need to free up the reserve
1289 * block pool so that the used block count in the superblock on
1290 * disk is correct at the end of the remount. Stash the current
1291 * reserve pool size so that if we get remounted rw, we can
1292 * return it to the same size.
1294 xfs_save_resvblks(mp);
1295 xfs_quiesce_attr(mp);
1296 mp->m_flags |= XFS_MOUNT_RDONLY;
1303 * Second stage of a freeze. The data is already frozen so we only
1304 * need to take care of the metadata. Once that's done sync the superblock
1305 * to the log to dirty it in case of a crash while frozen. This ensures that we
1306 * will recover the unlinked inode lists on the next mount.
1310 struct super_block *sb)
1312 struct xfs_mount *mp = XFS_M(sb);
1314 xfs_save_resvblks(mp);
1315 xfs_quiesce_attr(mp);
1316 return xfs_sync_sb(mp, true);
1321 struct super_block *sb)
1323 struct xfs_mount *mp = XFS_M(sb);
1325 xfs_restore_resvblks(mp);
1326 xfs_log_work_queue(mp);
1331 xfs_fs_show_options(
1333 struct dentry *root)
1335 return xfs_showargs(XFS_M(root->d_sb), m);
1339 * This function fills in xfs_mount_t fields based on mount args.
1340 * Note: the superblock _has_ now been read in.
1344 struct xfs_mount *mp)
1346 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1348 /* Fail a mount where the logbuf is smaller than the log stripe */
1349 if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1350 if (mp->m_logbsize <= 0 &&
1351 mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) {
1352 mp->m_logbsize = mp->m_sb.sb_logsunit;
1353 } else if (mp->m_logbsize > 0 &&
1354 mp->m_logbsize < mp->m_sb.sb_logsunit) {
1356 "logbuf size must be greater than or equal to log stripe size");
1360 /* Fail a mount if the logbuf is larger than 32K */
1361 if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) {
1363 "logbuf size for version 1 logs must be 16K or 32K");
1369 * V5 filesystems always use attr2 format for attributes.
1371 if (xfs_sb_version_hascrc(&mp->m_sb) &&
1372 (mp->m_flags & XFS_MOUNT_NOATTR2)) {
1374 "Cannot mount a V5 filesystem as %s. %s is always enabled for V5 filesystems.",
1375 MNTOPT_NOATTR2, MNTOPT_ATTR2);
1380 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1381 * told by noattr2 to turn it off
1383 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1384 !(mp->m_flags & XFS_MOUNT_NOATTR2))
1385 mp->m_flags |= XFS_MOUNT_ATTR2;
1388 * prohibit r/w mounts of read-only filesystems
1390 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1392 "cannot mount a read-only filesystem as read-write");
1396 if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) &&
1397 (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE)) &&
1398 !xfs_sb_version_has_pquotino(&mp->m_sb)) {
1400 "Super block does not support project and group quota together");
1409 struct super_block *sb,
1414 struct xfs_mount *mp = NULL;
1415 int flags = 0, error = -ENOMEM;
1417 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
1421 spin_lock_init(&mp->m_sb_lock);
1422 mutex_init(&mp->m_growlock);
1423 atomic_set(&mp->m_active_trans, 0);
1424 INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker);
1425 INIT_DELAYED_WORK(&mp->m_eofblocks_work, xfs_eofblocks_worker);
1426 mp->m_kobj.kobject.kset = xfs_kset;
1431 error = xfs_parseargs(mp, (char *)data);
1433 goto out_free_fsname;
1435 sb_min_blocksize(sb, BBSIZE);
1436 sb->s_xattr = xfs_xattr_handlers;
1437 sb->s_export_op = &xfs_export_operations;
1438 #ifdef CONFIG_XFS_QUOTA
1439 sb->s_qcop = &xfs_quotactl_operations;
1440 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
1442 sb->s_op = &xfs_super_operations;
1445 flags |= XFS_MFSI_QUIET;
1447 error = xfs_open_devices(mp);
1449 goto out_free_fsname;
1451 error = xfs_init_mount_workqueues(mp);
1453 goto out_close_devices;
1455 error = xfs_icsb_init_counters(mp);
1457 goto out_destroy_workqueues;
1459 error = xfs_readsb(mp, flags);
1461 goto out_destroy_counters;
1463 error = xfs_finish_flags(mp);
1467 error = xfs_setup_devices(mp);
1471 error = xfs_filestream_mount(mp);
1476 * we must configure the block size in the superblock before we run the
1477 * full mount process as the mount process can lookup and cache inodes.
1479 sb->s_magic = XFS_SB_MAGIC;
1480 sb->s_blocksize = mp->m_sb.sb_blocksize;
1481 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1482 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1483 sb->s_max_links = XFS_MAXLINK;
1484 sb->s_time_gran = 1;
1485 set_posix_acl_flag(sb);
1487 /* version 5 superblocks support inode version counters. */
1488 if (XFS_SB_VERSION_NUM(&mp->m_sb) == XFS_SB_VERSION_5)
1489 sb->s_flags |= MS_I_VERSION;
1491 error = xfs_mountfs(mp);
1493 goto out_filestream_unmount;
1495 root = igrab(VFS_I(mp->m_rootip));
1500 sb->s_root = d_make_root(root);
1508 out_filestream_unmount:
1509 xfs_filestream_unmount(mp);
1512 out_destroy_counters:
1513 xfs_icsb_destroy_counters(mp);
1514 out_destroy_workqueues:
1515 xfs_destroy_mount_workqueues(mp);
1517 xfs_close_devices(mp);
1519 xfs_free_fsname(mp);
1525 xfs_filestream_unmount(mp);
1530 STATIC struct dentry *
1532 struct file_system_type *fs_type,
1534 const char *dev_name,
1537 return mount_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super);
1541 xfs_fs_nr_cached_objects(
1542 struct super_block *sb,
1543 struct shrink_control *sc)
1545 return xfs_reclaim_inodes_count(XFS_M(sb));
1549 xfs_fs_free_cached_objects(
1550 struct super_block *sb,
1551 struct shrink_control *sc)
1553 return xfs_reclaim_inodes_nr(XFS_M(sb), sc->nr_to_scan);
1556 static const struct super_operations xfs_super_operations = {
1557 .alloc_inode = xfs_fs_alloc_inode,
1558 .destroy_inode = xfs_fs_destroy_inode,
1559 .evict_inode = xfs_fs_evict_inode,
1560 .drop_inode = xfs_fs_drop_inode,
1561 .put_super = xfs_fs_put_super,
1562 .sync_fs = xfs_fs_sync_fs,
1563 .freeze_fs = xfs_fs_freeze,
1564 .unfreeze_fs = xfs_fs_unfreeze,
1565 .statfs = xfs_fs_statfs,
1566 .remount_fs = xfs_fs_remount,
1567 .show_options = xfs_fs_show_options,
1568 .nr_cached_objects = xfs_fs_nr_cached_objects,
1569 .free_cached_objects = xfs_fs_free_cached_objects,
1572 static struct file_system_type xfs_fs_type = {
1573 .owner = THIS_MODULE,
1575 .mount = xfs_fs_mount,
1576 .kill_sb = kill_block_super,
1577 .fs_flags = FS_REQUIRES_DEV,
1579 MODULE_ALIAS_FS("xfs");
1582 xfs_init_zones(void)
1585 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
1586 if (!xfs_ioend_zone)
1589 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
1591 if (!xfs_ioend_pool)
1592 goto out_destroy_ioend_zone;
1594 xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
1596 if (!xfs_log_ticket_zone)
1597 goto out_destroy_ioend_pool;
1599 xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t),
1600 "xfs_bmap_free_item");
1601 if (!xfs_bmap_free_item_zone)
1602 goto out_destroy_log_ticket_zone;
1604 xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
1606 if (!xfs_btree_cur_zone)
1607 goto out_destroy_bmap_free_item_zone;
1609 xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t),
1611 if (!xfs_da_state_zone)
1612 goto out_destroy_btree_cur_zone;
1614 xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
1615 if (!xfs_ifork_zone)
1616 goto out_destroy_da_state_zone;
1618 xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
1619 if (!xfs_trans_zone)
1620 goto out_destroy_ifork_zone;
1622 xfs_log_item_desc_zone =
1623 kmem_zone_init(sizeof(struct xfs_log_item_desc),
1624 "xfs_log_item_desc");
1625 if (!xfs_log_item_desc_zone)
1626 goto out_destroy_trans_zone;
1629 * The size of the zone allocated buf log item is the maximum
1630 * size possible under XFS. This wastes a little bit of memory,
1631 * but it is much faster.
1633 xfs_buf_item_zone = kmem_zone_init(sizeof(struct xfs_buf_log_item),
1635 if (!xfs_buf_item_zone)
1636 goto out_destroy_log_item_desc_zone;
1638 xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) +
1639 ((XFS_EFD_MAX_FAST_EXTENTS - 1) *
1640 sizeof(xfs_extent_t))), "xfs_efd_item");
1642 goto out_destroy_buf_item_zone;
1644 xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) +
1645 ((XFS_EFI_MAX_FAST_EXTENTS - 1) *
1646 sizeof(xfs_extent_t))), "xfs_efi_item");
1648 goto out_destroy_efd_zone;
1651 kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
1652 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | KM_ZONE_SPREAD,
1653 xfs_fs_inode_init_once);
1654 if (!xfs_inode_zone)
1655 goto out_destroy_efi_zone;
1658 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
1659 KM_ZONE_SPREAD, NULL);
1661 goto out_destroy_inode_zone;
1662 xfs_icreate_zone = kmem_zone_init(sizeof(struct xfs_icreate_item),
1664 if (!xfs_icreate_zone)
1665 goto out_destroy_ili_zone;
1669 out_destroy_ili_zone:
1670 kmem_zone_destroy(xfs_ili_zone);
1671 out_destroy_inode_zone:
1672 kmem_zone_destroy(xfs_inode_zone);
1673 out_destroy_efi_zone:
1674 kmem_zone_destroy(xfs_efi_zone);
1675 out_destroy_efd_zone:
1676 kmem_zone_destroy(xfs_efd_zone);
1677 out_destroy_buf_item_zone:
1678 kmem_zone_destroy(xfs_buf_item_zone);
1679 out_destroy_log_item_desc_zone:
1680 kmem_zone_destroy(xfs_log_item_desc_zone);
1681 out_destroy_trans_zone:
1682 kmem_zone_destroy(xfs_trans_zone);
1683 out_destroy_ifork_zone:
1684 kmem_zone_destroy(xfs_ifork_zone);
1685 out_destroy_da_state_zone:
1686 kmem_zone_destroy(xfs_da_state_zone);
1687 out_destroy_btree_cur_zone:
1688 kmem_zone_destroy(xfs_btree_cur_zone);
1689 out_destroy_bmap_free_item_zone:
1690 kmem_zone_destroy(xfs_bmap_free_item_zone);
1691 out_destroy_log_ticket_zone:
1692 kmem_zone_destroy(xfs_log_ticket_zone);
1693 out_destroy_ioend_pool:
1694 mempool_destroy(xfs_ioend_pool);
1695 out_destroy_ioend_zone:
1696 kmem_zone_destroy(xfs_ioend_zone);
1702 xfs_destroy_zones(void)
1705 * Make sure all delayed rcu free are flushed before we
1709 kmem_zone_destroy(xfs_icreate_zone);
1710 kmem_zone_destroy(xfs_ili_zone);
1711 kmem_zone_destroy(xfs_inode_zone);
1712 kmem_zone_destroy(xfs_efi_zone);
1713 kmem_zone_destroy(xfs_efd_zone);
1714 kmem_zone_destroy(xfs_buf_item_zone);
1715 kmem_zone_destroy(xfs_log_item_desc_zone);
1716 kmem_zone_destroy(xfs_trans_zone);
1717 kmem_zone_destroy(xfs_ifork_zone);
1718 kmem_zone_destroy(xfs_da_state_zone);
1719 kmem_zone_destroy(xfs_btree_cur_zone);
1720 kmem_zone_destroy(xfs_bmap_free_item_zone);
1721 kmem_zone_destroy(xfs_log_ticket_zone);
1722 mempool_destroy(xfs_ioend_pool);
1723 kmem_zone_destroy(xfs_ioend_zone);
1728 xfs_init_workqueues(void)
1731 * The allocation workqueue can be used in memory reclaim situations
1732 * (writepage path), and parallelism is only limited by the number of
1733 * AGs in all the filesystems mounted. Hence use the default large
1734 * max_active value for this workqueue.
1736 xfs_alloc_wq = alloc_workqueue("xfsalloc",
1737 WQ_MEM_RECLAIM|WQ_FREEZABLE, 0);
1745 xfs_destroy_workqueues(void)
1747 destroy_workqueue(xfs_alloc_wq);
1755 printk(KERN_INFO XFS_VERSION_STRING " with "
1756 XFS_BUILD_OPTIONS " enabled\n");
1760 error = xfs_init_zones();
1764 error = xfs_init_workqueues();
1766 goto out_destroy_zones;
1768 error = xfs_mru_cache_init();
1770 goto out_destroy_wq;
1772 error = xfs_buf_init();
1774 goto out_mru_cache_uninit;
1776 error = xfs_init_procfs();
1778 goto out_buf_terminate;
1780 error = xfs_sysctl_register();
1782 goto out_cleanup_procfs;
1784 xfs_kset = kset_create_and_add("xfs", NULL, fs_kobj);
1787 goto out_sysctl_unregister;;
1791 xfs_dbg_kobj.kobject.kset = xfs_kset;
1792 error = xfs_sysfs_init(&xfs_dbg_kobj, &xfs_dbg_ktype, NULL, "debug");
1794 goto out_kset_unregister;
1797 error = xfs_qm_init();
1799 goto out_remove_kobj;
1801 error = register_filesystem(&xfs_fs_type);
1810 xfs_sysfs_del(&xfs_dbg_kobj);
1811 out_kset_unregister:
1813 kset_unregister(xfs_kset);
1814 out_sysctl_unregister:
1815 xfs_sysctl_unregister();
1817 xfs_cleanup_procfs();
1819 xfs_buf_terminate();
1820 out_mru_cache_uninit:
1821 xfs_mru_cache_uninit();
1823 xfs_destroy_workqueues();
1825 xfs_destroy_zones();
1834 unregister_filesystem(&xfs_fs_type);
1836 xfs_sysfs_del(&xfs_dbg_kobj);
1838 kset_unregister(xfs_kset);
1839 xfs_sysctl_unregister();
1840 xfs_cleanup_procfs();
1841 xfs_buf_terminate();
1842 xfs_mru_cache_uninit();
1843 xfs_destroy_workqueues();
1844 xfs_destroy_zones();
1847 module_init(init_xfs_fs);
1848 module_exit(exit_xfs_fs);
1850 MODULE_AUTHOR("Silicon Graphics, Inc.");
1851 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
1852 MODULE_LICENSE("GPL");